Re: [HELP]sqlldr question

Hantie schrieb:
> by the way, under a scenerio is that.. everytime sqlldr will load data
> records no more than 1 million.
> Can I issue a command like below seems will have a (all or none) insertion
> result:
>
> sqlldr ***/**** errors=0 rows=1000000 *********
>
> what do u think?
>
>
> "Hantie" <blk330ci_at_hotmail.com> 撰寫於郵件新聞
> :d86rfe$jf71_at_imsp212.netvigator.com...
>

>>thanks Maxim..
>>I've tested both Direct/conventional path and subsequently all records

>>commit before error occurs.
>>it seems to me that there is noway to make sqlldr rollback, am i correct?
>>
>>Jan
>>
>>"Maxim Demenko" <mdemenko_at_gmail.com> 撰寫於郵件新聞
>>:d851sq$tfc$02$1_at_news.t-online.com...
>>
>>>"If you use direct path, commit happens at the end of the load," --
>>>Maxim Demenko , "Re: [HELP]sqlldr question" (21:36)
>>>
>>>Sorry , it was not quite correct, during the direct load happens data
>>>save , so if error occurs at the end of load , loaded records could not
>>>be rolled back as well.
>>>
>>>Best regards
>>>
>>>Maxim
>>
>>

Well, the documentation states (for conventional path) that ROWS multiplied with the row length must fit into BINDSIZE ( if you not specify ROWS, bind array will be filled with rows until it possible), BINDSIZE in turn depends on the READSIZE and can not be greater than that , READSIZE is limited ( for 9i ) by 20 Mb. Of course you can play with these parameters to delay the point of commit as much as possible, but in my opinion the better way is suggested by Chris - to load into a staging table and check , whether errors had place.

Best regards

Maxim Received on Wed Jun 08 2005 - 13:13:35 CDT